cpu/stm32: add support for STM32C0 and NUCLEO-C031C6

[jparker324]

Cleaned-up commit history of PR #20267

Support added for ST Micro's board NUCLEO-C031C6, it's cpu STM32C031C6, and the rest of the cpu family STM32C0XX (as CPU_FAM_STM32C0). This currently includes STM32C011 and STM32C031, but code for identification has also been added for the STM32C071 and STM32C091 lines due out later this year.

Code was adapted from CPU_FAM_STM32G0 where applicable.

The most important difference in the STM32C0 product line is the removal of the PLL. As coded the default clock source is HSI.

fixes #19210

[benpicco]

Code looks good to me!
Nice addition.

[riot-ci]

Murdock results

✔️ PASSED

a67f778 tests: add nucleo-c031c6 to Makefile.ci

Success	Failures	Total	Runtime
8629	0	8629	08m:30s

Artifacts

• Documentation preview

[benpicco]

One more thing: Since the STM32C031C6 has very little memory (32k ROM, 12k RAM) I expect that some applications won't fit on it.

CI tries to build every application for every board, so you need to tell it which ones to exclude for a board because they won't fit. You can do this by running

dist/tools/insufficient_memory/update_insufficient_memory_board.sh nucleo-c031c6

This will build all applications & tests for nucleo-c031c6 and update their Makefile.ci if the application does not fit.

You then need to commit the updated Makefile.ci files.

[benpicco]

Looks like there is no low power timer on STM32C0 - so you might have to remove the periph_rtt feature.
Or is there a timer that keeps running in Deep Sleep?

[jparker324]

There is not a dedicated low-power timer LPTIM peripheral like on previous ST parts (example STM32G071 section 3.18.4). Though they offer, according to datasheet STM32C031x4/x6 section 3.15 and 3.16, the LSI as a low-power RC clock source and there are 2 low-power timers that support a couple low-power RTC interrupts. I don't know if that meets all the requirements to support RIOT's periph_rtt.

[benpicco]

The only requirement for periph_rtt is that it keeps ticking in the deepest sleep state (and can wake up/reset the CPU).
But you can also provide that as a follow-up PR and just drop advertising the feature for now.

[jparker324]

Alright. I'll do that. I still get build errors for tests/pkg/lvgl, tests/pkg/lvgl_touch, tests/sys/progress_bar, and tests/sys/xtimer_hang. lvgl and lvgl_touch also fail when I try them under nucleo_g071rb for example, so that's either expected or local to my environment. The progress_bar and xtimer_hang though are "OK" for that board.

[jparker324]

I think the failures of tests/sys/progress_bar and tests/sys/xtimer_hang were due to previous interrupted or cancelled builds. I cleaned their build and bin dirs, and they both appear to build fine now.

tests/sys/progress_bar OK
tests/sys/xtimer_hang OK

LVGL still fails though. I also get a brief message on spi_dma before it's not supported output, although it's not explicitly build failed.

tests/periph/spi_dma                    dist/tools/insufficient_memory/../../../tests/periph/spi_dma/Makefile.ci:3: ../spi/Makefile.ci: No such file or directory
make: *** No rule to make target '../spi/Makefile.ci'.  Stop.
not supported

[benpicco]

I also get a brief message on spi_dma before it's not supported output, although it's not explicitly build failed.

That's because you don't advertise the periph_dma feature, but as you also didn't configure any DMA channels for the SPI, that's correct.

[jparker324]

I've restored and re-built the Makefile.ci files again. They look better now. If I had to guess, it was probably another result of earlier cancelled builds.

[benpicco]

looks like (at least) tests/pkg/lvgl_touch also needs a Makefile.ci update

[jparker324]

1. Does anything need to be done about the tests/sys/ztimer_overhead error yesterday?

make: Entering directory '/tmp/dwq.0.29611624904775513/5b84c211eba9dea6bec9bf777727be8b/tests/sys/ztimer_overhead'
r
/tmp/dwq.0.29611624904775513/5b84c211eba9dea6bec9bf777727be8b/build/tests_ztimer_overhead.elf  tap0 
RIOT native interrupts/signals initialized.
RIOT native board initialized.
RIOT native hardware initialization complete.

Help: Press s to start test, r to print it is ready
READY
s
START
main(): This is RIOT! (Version: buildtest)
ZTIMER_USEC auto_adjust params:
    ZTIMER_USEC->adjust_set = 11
    ZTIMER_USEC->adjust_sleep = 5
ZTIMER_USEC auto_adjust params cleared
zitmer_overhead_set...
min=7 max=43 avg_diff=13
zitmer_overhead_sleep...
min=11 max=255 avg_diff=36
ZTIMER_USEC adjust params for native:
    CONFIG_ZTIMER_USEC_ADJUST_SET    7
    CONFIG_ZTIMER_USEC_ADJUST_SLEEP  11

make: Leaving directory '/tmp/dwq.0.29611624904775513/5b84c211eba9dea6bec9bf777727be8b/tests/sys/ztimer_overhead'
Traceback (most recent call last):
  File "/tmp/dwq.0.29611624904775513/5b84c211eba9dea6bec9bf777727be8b/tests/sys/ztimer_overhead/tests/01-run.py", line 33, in <module>
    sys.exit(run(testfunc))
  File "/tmp/dwq.0.29611624904775513/5b84c211eba9dea6bec9bf777727be8b/dist/pythonlibs/testrunner/__init__.py", line 30, in run
    testfunc(child)
  File "/tmp/dwq.0.29611624904775513/5b84c211eba9dea6bec9bf777727be8b/tests/sys/ztimer_overhead/tests/01-run.py", line 29, in testfunc
    assert auto_adjust_sleep >= adjust_sleep - ADJUST_SLEEP_MARGIN
AssertionError
make: *** [/tmp/dwq.0.29611624904775513/5b84c211eba9dea6bec9bf777727be8b/makefiles/tests/tests.inc.mk:26: test] Error 1

2. When running ./dist/tools/insufficient_memory/update_insufficient_memory_board.sh nucleo-c031c6 do you not see the error in the lvgl and lvgl_touch packages? How did it/I miss the other Makefile.ci you corrected?

tests/pkg/lvgl                          build failed
make: Entering directory '/home/parkerj/RIOT/tests/pkg/lvgl'
Launching build container using image "docker.io/riot/riotbuild:latest".
docker run --rm --tty --user $(id -u) -v '/usr/share/zoneinfo/America/New_York:/etc/localtime:ro' -v '/home/parkerj/RIOT:/data/riotbuild/riotbase:delegated' -v '/home/parkerj/.cargo/registry:/data/riotbuild/.cargo/registry:delegated' -v '/home/parkerj/.cargo/git:/data/riotbuild/.cargo/git:delegated' -e 'RIOTBASE=/data/riotbuild/riotbase' -e 'CCACHE_BASEDIR=/data/riotbuild/riotbase' -e 'BUILD_DIR=/data/riotbuild/riotbase/build' -e 'RIOTPROJECT=/data/riotbuild/riotbase' -e 'RIOTCPU=/data/riotbuild/riotbase/cpu' -e 'RIOTBOARD=/data/riotbuild/riotbase/boards' -e 'RIOTMAKE=/data/riotbuild/riotbase/makefiles'      -e 'BOARD=nucleo-c031c6' -e 'RIOT_CI_BUILD=1' -e 'DISABLE_MODULE=test_utils_interactive_sync' -e 'DEFAULT_MODULE=' -e 'FEATURES_REQUIRED=' -e 'FEATURES_BLACKLIST=' -e 'FEATURES_OPTIONAL=' -e 'USEMODULE=lvgl_contrib lvgl_extra_layout_flex lvgl_extra_theme_default lvgl_extra_theme_default_dark lvgl_extra_widget_chart lvgl_extra_widget_win' -e 'USEPKG=lvgl'  -w '/data/riotbuild/riotbase/tests/pkg/lvgl/' 'docker.io/riot/riotbuild:latest' make 'BOARD=nucleo-c031c6'   -j12  
Building application "tests_lvgl" for "nucleo-c031c6" with MCU "stm32".

/tmp/ccXrsj5w.s: Assembler messages:
/tmp/ccXrsj5w.s: Error: unaligned opcodes detected in executable segment

[benpicco]

That's just a sporadic failure that occurs when CI is under high load.
Re-running the build usually fixes it.

[benpicco]

That was a great first PR!
Let's continue with GD32E23x 😃

[jparker324]

Thanks for your help. I'll get back to GD32 it but it might be a few weeks, yet.